Digital Video Broadcasting (DVB); Subtitling systems

Transcription

1 !!!!!!!!!!!!!! Digital Video Broadcasting (DVB); Subtitling systems DVB Document A009! July 2013

2

3 3 Contents Intellectual Property Rights... 5 Foreword Scope References Normative references Informative references Definitions and abbreviations Definitions Abbreviations Introduction to DVB subtitling system Overview Data hierarchy and terminology Temporal hierarchy and terminology Subtitle decoder model Decoder temporal model Service acquisition Presentation Time Stamps (PTS) Display definition Page composition Region composition Points to note Buffer memory model Pixel buffer memory Region memory Composition buffer memory Cumulative display construction Decoder rendering bandwidth model Page erasure Region move or change in visibility Region fill CLUT modification Graphic object decoding Character object decoding PES packet format The PES packet data for subtitling Syntax and semantics of the PES data field for subtitling Syntax and semantics of the subtitling segment Display definition segment Page composition segment Region composition segment CLUT definition segment Object data segment Pixel-data sub-block Syntax and semantics of the pixel code strings End of display set segment Disparity Signalling Segment Requirements for the subtitling data Scope of Identifiers Scope of dependencies Composition page Ancillary page... 39

4 4 8.3 Order of delivery PTS field Positioning of regions and objects Regions Objects sharing a PTS Objects added to a region Translation to colour components to 2-bit reduction to 2-bit reduction to 4-bit reduction Default CLUTs and map-tables contents entry CLUT default contents entry CLUT default contents entry CLUT default contents _to_4-bit_map-table default contents _to_8-bit_map-table default contents _to_8-bit_map-table default contents Structure of the pixel code strings (informative) Annex A (informative): How the DVB subtitling system works A.1 Data hierarchy and terminology A.2 Temporal hierarchy and terminology A.3 Decoder temporal model A.4 Decoder display technology model A.4.1 Region based with indexed colours A.4.2 Colour quantization A.5 Examples of the subtitling system in operation A.5.1 Double buffering A Instant graphics A Stenographic subtitles Annex B (informative): Use of the DDS for HDTV services Annex C (informative): Illustration of the application of the disparity_shift_update_sequence mechanism for 3D content Annex D (informative): Bibliography Annex E (Informative): General Guidelines Subregions Disparity Updates Timing 59 Legacy receivers History... 63

5 5 Intellectual Property Rights IPRs essential or potentially essential to the present document may have been declared to ETSI. The information pertaining to these essential IPRs, if any, is publicly available for ETSI members and non-members, and can be found in ETSI SR : "Intellectual Property Rights (IPRs); Essential, or potentially Essential, IPRs notified to ETSI in respect of ETSI standards", which is available from the ETSI Secretariat. Latest updates are available on the ETSI Web server ( Pursuant to the ETSI IPR Policy, no investigation, including IPR searches, has been carried out by ETSI. No guarantee can be given as to the existence of other IPRs not referenced in ETSI SR (or the updates on the ETSI Web server) which are, or may be, or may become, essential to the present document. Foreword This European Standard (EN) has been produced by Joint Technical Committee (JTC) Broadcast of the European Broadcasting Union (EBU), Comité Européen de Normalisation ELECtrotechnique (CENELEC) and the European Telecommunications Standards Institute (ETSI). NOTE: The EBU/ETSI JTC Broadcast was established in 1990 to co-ordinate the drafting of standards in the specific field of broadcasting and related fields. Since 1995 the JTC Broadcast became a tripartite body by including in the Memorandum of Understanding also CENELEC, which is responsible for the standardization of radio and television receivers. The EBU is a professional association of broadcasting organizations whose work includes the co-ordination of its members' activities in the technical, legal, programme-making and programme-exchange domains. The EBU has active members in about 60 countries in the European broadcasting area; its headquarters is in Geneva. European Broadcasting Union CH-1218 GRAND SACONNEX (Geneva) Switzerland Tel: Fax: The Digital Video Broadcasting Project (DVB) is an industry-led consortium of broadcasters, manufacturers, network operators, software developers, regulatory bodies, content owners and others committed to designing global standards for the delivery of digital television and data services. DVB fosters market driven solutions that meet the needs and economic circumstances of broadcast industry stakeholders and consumers. DVB standards cover all aspects of digital television from transmission through interfacing, conditional access and interactivity for digital video, audio and data. The consortium came together in 1993 to provide global standardisation, interoperability and future proof specifications. National transposition dates Date of adoption of this EN: 29 September 2011 Date of latest announcement of this EN (doa): 31 December 2011 Date of latest publication of new National Standard or endorsement of this EN (dop/e): 30 June 2012 Date of withdrawal of any conflicting National Standard (dow): 30 June 2012

6 6 1 Scope The present document specifies the method by which subtitles, logos and other graphical elements may be coded and carried in DVB bitstreams. The system applies Colour Look-Up Tables (CLUTs) to define the colours of the graphical elements. The transport of the coded graphical elements is based on the MPEG-2 system described in ISO/IEC [1]. 2 References References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the reference document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found at NOTE: While any hyperlinks included in this clause were valid at the time of publication ETSI cannot guarantee their long term validity. 2.1 Normative references The following referenced documents are necessary for the application of the present document. [1] ISO/IEC : "Information technology - Generic coding of moving pictures and associated audio information: Systems". [2] ETSI EN : "Digital Video Broadcasting (DVB); Specification for Service Information (SI) in DVB systems". [3] ITU-R Recommendation BT.601: "Studio encoding parameters of digital television for standard 4:3 and wide-screen 16:9 aspect ratios". [4] ITU-R Recommendation BT.656-4: "Interfaces for digital component video signals in 525-line and 625-line television systems operating at the 4:2:2 level of Recommendation ITU-R BT.601 (Part A)". [5] ETSI EN (V1.2.1): "Digital Video Broadcasting (DVB); Subtitling systems". [6] ETSI EN (V1.3.1): "Digital Video Broadcasting (DVB); Subtitling systems". 2.2 Informative references The following referenced documents are not necessary for the application of the present document but they assist the user with regard to a particular subject area. Not applicable.

7 7 3 Definitions and abbreviations 3.1 Definitions For the purposes of the present document, the following terms and definitions apply: ancillary page: means of conveying subtitle elements that may be shared by multiple subtitle services within a subtitle stream NOTE: For example, an ancillary page can be used to carry logos or character glyphs. Colour Look-Up Table (CLUT): look-up table applied in each region for translating the objects' pseudo-colours into the correct colours to be displayed CLUT-family: family of CLUTs which may consist of: one CLUT with 4 entries; one CLUT with 16 entries; one CLUT with 256 entries. A CLUT-family is used in a region to define colours for decoders with different rendering capabilities. NOTE: Three CLUTs are defined to allow flexibility in the decoder design. Not all decoders may support a CLUT with 256 entries, some may provide sixteen or even only four entries. A palette of four colours might be enough for graphics that are basically monochrome, like very simple subtitles, while a palette of sixteen colours allows for cartoon-like coloured objects or coloured subtitles with antialiassed edges. composition page: means of conveying subtitle elements for one specific subtitle service display definition: definition of the video image display size for which a subtitle stream has been prepared display set: set of subtitle segments of a specific subtitle service to which the same PTS value is associated epoch: period of time for which the decoder maintains an invariant memory layout NOTE: This layout may be altered by resets to the decoder state caused by receiving page composition segments with page state = "mode change". The end of an epoch therefore signals the "death" of a page. The epoch may, if so desired, be considered to be the highest level data structure in DVB subtitling. object: graphical unit that can be positioned within a region; examples of an object include a character glyph, a logo, a map, etc. NOTE: Each object has its own object_id. Packet IDentifier (PID): Transport packet identifier NOTE: See ISO/IEC [1]. page: set of subtitles for a subtitle service during a certain period NOTE: A page consists of one or more page instances. Each page update or refresh will result in a new page instance. A page contains a number of regions, and in each region there may be a number of objects. page composition: composition (use and positioning) of regions that may be displayed within the page NOTE: At any new page instance the page composition may change; for example, some regions may not yet or no longer be displayed. At any one time, only one page composition can be active for displaying. page instance: period of time during which that page does not change i.e. there is no change to the page composition, to any region composition, to any object within a region or any applicable CLUT NOTE: Typically, a new page instance is defined by the PTS of a display set.

8 8 PES packet: See ISO/IEC [1]. pixel-data: string of data bytes that contains, in coded form, the representation of a graphical object Presentation Time Stamp (PTS): See ISO/IEC [1]. region: rectangular area on the page in which objects can be positioned NOTE: Regions may be shared by multiple subtitling services within the same subtitle stream. Objects that share one or more horizontal scan lines on the screen are included in the same region. region composition: composition (use and positioning) of objects within a region subtitle element: subtitle data used within a page composition and contained within a subtitle segment NOTE: Regions, region compositions, CLUTs and object data are examples of subtitle elements. subtitle segment: basic syntactical element of a subtitle stream subtitle service: service that provides subtitling for a program for a certain purpose, such as subtitles in a specific language or for the hard of hearing NOTE 1: A subtitle service is displayed as a series of one or more pages. NOTE 2: Typically, a subtitle service meets a single communication requirement (e.g. the graphics to provide subtitles in one language for one program). subtitle stream: stream of subtitling segments carried in transport packets identified by the same PID NOTE: A subtitle stream contains one or more subtitle services. transport packet: See ISO/IEC [1]. transport stream: stream of transport packets carrying one or more MPEG programs NOTE: See ISO/IEC [1]. 3.2 Abbreviations For the purposes of the present document, the following abbreviations apply: B bslbf Cb Blue value of colour representation in default CLUT bit string, left bit first Chrominance value representing B-Y colour difference signal NOTE: As defined in ITU-R Recommendation BT.601 [3], clause CLUT Cr Colour Look-Up Table Chrominance value representing R-Y colour difference signal NOTE: As defined in ITU-R Recommendation BT.601 [3], clause DTV EDS EIT DDS DSS DVB G GOP HDTV IRD MPEG NOTE: Digital TeleVision End of Display Set Segment Event Information Table Display Definition Segment Disparity Signalling Segment Digital Video Broadcasting Green value of colour representation in default CLUT Group of Pictures High Definition TeleVision Integrated Receiver Decoder Moving Pictures Experts Group WG11 in SC 29 of JTC1 of ISO/IEC.

9 9 PCR PCS PES Programme Clock Reference Page Composition Segment Packetized Elementary Stream NOTE: As defined in ISO/IEC [1]. PID transport Packet IDentifier NOTE: As defined in ISO/IEC [1]. PMT Program Map Table NOTE: As defined in ISO/IEC [1]. PTS Presentation Time Stamp NOTE: As defined in ISO/IEC [1]. R RCS ROM SDT STC T TS Red value of colour representation in default CLUT Region Composition Segment Read-Only Memory Service Description Table System Time Clock Transparency value Transport Stream NOTE: As defined in ISO/IEC [1]. uimsbf tcimsbf Y unsigned integer, most significant bit first two's complement integer, msb (sign) bit first luminance value NOTE: As defined in ITU-R Recommendation BT.601 [3], clause Introduction to DVB subtitling system The present document specifies the DVB subtitling system for the transport and coding of subtitles. 4.1 Overview The DVB subtitling system defined in the present document provides a syntax for decoding subtitle streams. A subtitle stream conveys one or more subtitle services; each service containing the textual and/or graphical information needed to provide subtitles or glyphs for a particular purpose. Separate subtitle services may be used, for example, to convey subtitles in several languages. Each subtitle service displays its information in a sequence of so-called pages that are intended to be overlayed on the associated video image. A subtitle page contains one or more regions, each region being a rectangular area with a specified set of attributes. These attributes include a region identifier, the horizontal and vertical size, pixel depth and background colour. A region is used as the background structure into which graphical objects are placed. An object may represent a character, a word, a line of text or an entire sentence; it might also define a logo or icon. The use and positioning of objects within a region is defined by the region composition segment. The use and positioning of regions within a page is defined by the page composition segment, in which a list of displayed regions is provided, each with their own spatial position. A page composition need not change when objects are added to or removed from a region. Furthermore regions may be declared but not used. By way of example one region can be used to display multiple subtitle fragments, as depicted in figure 1. First the text "Shall we?" is displayed in the region; subsequently this text is removed and the new text "Said the fly on the mirror" is displayed. It is possible to use more than one region at the same time; for example one region could be used to display subtitles on the bottom of the screen, while another one might be used to display a logo somewhere else on the screen.

10 10 Region 1 Shall we? Region 2 Said the fly on the mirror. NOTE: The subtitles are positioned within the same region. Figure 1: Two regions overlayed on top of video; one with a logo and another one with subtitles A DVB subtitle stream is carried in PES packets and the timing of their presentation is defined by the PTS in the PES header. Upon reception and decoding of the subtitle data for a page (such as the page composition, the region composition, the objects to be used and any other associated data) the page contents are displayed at the time indicated by the associated PTS. When objects are to be added, the decoder receives region composition updates and the data for the new objects, and will display the updated page at the time indicated by the new PTS. At the page update only page differences need be provided. To improve random access to DVB subtitling, a page refresh is also possible. At page refresh all the subtitling data needed to display a page is provided. Each page update or refresh will result in a new page instance. A page ceases to exist after the time-out of the page, or when a new page is defined. To provide efficient use of display memory in the decoder the DVB subtitling system uses region based graphics with indexed pixel colours. Pixel depths of 2, 4 and 8 bits are supported allowing up to 4, 16 or 256 different pixel codes to be used in each region. Each region is associated with a single CLUT family to define the colour and transparency for each of the pixel codes. In most cases, one CLUT is sufficient to present correctly the colours of all objects in a region, but if it is not enough, the objects can be split horizontally into smaller objects across separate vertically adjacent regions with one CLUT each. The use of CLUTs allows colour schemes to be dynamic. The colours that correspond to the entries within the region can be redefined at any suitable time, for instance in case of a CLUT with four entries from a black-grey-white scheme to a blue-grey-yellow scheme. Furthermore, a graphical unit may be divided into several regions each using a different CLUT, i.e. a different colour scheme may be applied in each of the regions. At the discretion of the encoder, objects designed for displays supporting 16 or 256 colours can be decoded into displays supporting fewer colours. A quantization algorithm is defined to ensure that the result of this process can be predicted by the originator. Use of this feature allows a single data stream to be decoded by a population of decoders with mixed, and possibly evolving, capabilities. A subtitle stream may transport multiple subtitle service components. In this case the pages of one particular subtitle service are all identified by the same page-id value. This value is used when transporting the subtitling data so as to provide a mechanism to retrieve the data that is specific to a service from a subtitle stream. The subtitling system allows sharing of subtitling data between services within the same subtitle stream. A frequent and often preferred method is to convey the distinct services in different streams on separate PIDs. In either case the appropriate PID, language and page-ids will be signalled in the Program Map Table (PMT) for the television service of interest (language and page-id in the subtitling descriptor defined in DVB-SI [2]). These two approaches are illustrated in figure 2.

11 11 PMT. PID = X language = spanish composition-page_id = 1 ancillary-page _id = 3 language = italian composition-page_id = 2 ancillary-page_id = 3 Subtitle descriptors associated with this subtitle stream in the PMT PES packet data PID = X Subtitle data signalled by page-id = 1 (spanish) Subtitle data signalled by page-id = 2 (italian) Subtitle data signalled by page-id = 3 (shared) a: Example of use of different page_ids to distinguish between different subtitle languages for the same service (shown with a shared ancillary page) PMT. PID = X language = spanish composition-page_id = 1 ancillary-page_id = 1 PID=Y language = italian composition-page_id = 1 ancillary-page_id = 1 Subtitle descriptors associated with these subtitle streams in the PMT PES packet data PID = X PES packet data PID = Y Subtitle data signalled by page_id = 1 (spanish) Subtitle data signalled by page_id = 1 (italian) b: Example of use of PIDs to distinguish between different subtitle languages for the same service (shown with no ancillary page) Figure 2: Example of two ways of conveying dual language subtitles (one using shared data) In summary, the DVB subtitling system provides a number of techniques that allow efficient transmission of the subtitling data: objects that occur more than once within a region need only be transmitted once, and then positioned multiple times within the region; objects used in more than one subtitle service need only be transmitted once; pixel data within objects are compressed using run-length coding; where the gamut of colours required for part of a graphical object is suitably limited, that part can be coded using a smaller number of bits per pixel and a map table. For example, an 8-bit per pixel graphical object may contain areas coded as 4 or 2-bits per pixel each preceded by a map table to map the 16 or 4 colours used onto the 256 colour set of the region. Similarly, a 4-bit per pixel object may contain areas coded as 2-bits per pixel;

12 12 colour definitions can be coded using either 16 or 32-bits per CLUT entry. This provides a trade-off between colour accuracy and transmission bandwidth; only those CLUT values to be used need be transmitted. The above features are fully supported within the DVB subtitling system. Subtitle streams intended for HDTV may include an optional data structure called the display_definition_segment which explicitly defines the display size for which that stream has been created. Subtitle streams associated with standard definition TV services need not include a display_definition_segment and may be encoded in accordance with EN (V1.2.1) [5]. Such streams will be nevertheless be decodable by decoders compliant with this and any later versions of EN Subtitle streams associated with standard definition TV services and intended to be decoded by decoders designed to EN (V1.2.1) [5] shall not include a display_definition_segment. In addition, functionality is provided to allow more efficient operation where there are private agreements between the data provider and the manufacturer of the decoder: objects resident in ROM in the decoder can be referenced; character codes, or strings of character codes, can be used instead of objects with the graphical representation of the character(s). This requires the decoder to be able to generate glyphs for these codes. The private agreements required to enable these features are beyond the scope the present document. 4.2 Data hierarchy and terminology The basic "building block" of a DVB subtitle stream is the subtitling segment. These segments are carried in PES packets, which are in turn carried by transport packets. The number of segments carried in a PES packet is only limited by the maximum length of a PES packet, as defined by ISO/IEC [1]. A subtitle stream shall be carried in transport packets identified by the same PID. A single subtitle stream can carry several different subtitle services. All the subtitling data required for a subtitle service shall be carried by a single subtitle stream. The different subtitle services can be subtitles in different languages for a common program. Alternatively, they could in principle be for different programs (provided that the programs share a common PCR). However a single subtitle stream shall not convey both a subtitle service which includes a display_definition_segment and one that does not; in this case the subtitle services must be carried in separate streams and on separate PIDs. Subtitle streams intended for HDTV services and which include a display_definition_segment are distinguished from those which are intended for standard definition services and have been coded in accordance with EN (V1.2.1) [5] by the use of HDTV-specific stream_content & component_type values in the DVB component descriptor signalled in the SDT and EIT for that service [2]. This provides a means whereby legacy SD-only decoders should ignore streams which include display_definition_segments. Different subtitle services can also be supplied to address different display characteristics or to address special needs. For instance: different subtitle services might be provided for 4:3 and 16:9 aspect ratio displays; subtitle services might be provided specifically for viewers with impaired hearing. These may include graphical representations of sounds. Within a subtitle stream, a page id value is assigned to each segment. Segments can either contain data specific for one subtitle service, or data that is to be shared by more than one subtitle service. The data for a subtitle service shall be carried in segments identified by at most two different page id values: one page id value signalling segments with data specific for that subtitle service; the use of this type of data is mandatory; one page id value signalling segments with data that may be shared by multiple subtitle services; the use of this type of data is optional.

13 13 For each subtitle service a subtitling_descriptor as defined in EN [2] signals the page id values of the segments needed to decode that subtitle service. The subtitling descriptor shall be included in the PMT of the program and shall be associated to the PID that conveys the subtitle stream. In the subtitling descriptor the page id of segments with data specific to that service is referred to as the composition page id, while the page id of segments with shared data is referred to as the ancillary page id. For example, the ancillary page id might signal segments carrying a logo that is common to subtitles in several different languages. The PTS in the PES packet header provides presentation timing information for the subtitling data, and is associated with the subtitle data in all segments carried in that PES packet. The PTS defines the time at which the associated decoded segments should be presented. This may include removal of subtitles, for example when an entire region is removed or when all objects in a region are removed. There may be two or more PES packets with the same PTS value, for example when it is not possible or desirable to include all segments associated to the same PTS in one PES packet. The complete set of segments of a subtitle service that are associated to the same PTS is referred to as a display set. The last segment of a display set shall be followed by an "end_of-display-set segment", which signals that no more subtitling data associated to a certain PTS is needed for that service before decoding can commence. The display sets shall be delivered in their correct presentation-order, and the PTSs of subsequent display sets shall differ by more than one video frame period. For carriage of multiple types of subtitling data, several segment types are defined, in particular: display definition segment; a subtitle service may be intended or have been prepared for display sizes other than full standard definition television (i.e. other than 720 pixels by 576 lines eg for HDTV). The optional display definition segment explicitly defines the display size for which that service has been created; page composition segment; the decoding of a subtitle service will typically result in the display of subsequent pages, each consisting of one or more regions; the page composition segment carries information on the page composition, such as the list of included regions, the spatial position of each region, some time-out information for the page and the state of the page; region composition segment; in each region typically one or more objects are positioned, while using one specific CLUT, identified by a CLUT-id; the region composition segment carries information on the region composition and on region attributes, such as the horizontal and vertical size, the background colour, the pixel depth of the region, which CLUT is used and a list of included objects with their position within the region; CLUT definition segment; the CLUT definition segment contains information on a specific CLUT, identified by a CLUT-id, such as the colours used for a CLUT entry; object data segment; the object data segment carries information on a specific object; there are two types of objects, graphical objects and text objects. An object data segment with a graphical object contains run-length encoded bitmap colours, while a text object carries a string of one character codes; end of display set segment; the end of display set segment contains no internal information, but is used to signal explicitly that no more segments need to be received before the decoding of the current display set can commence. The page id value of a segment containing data for a subtitle service shall be equal either to the value of the composition_page_id or the ancillary_page_id provided in the subtitle descriptor. Page compositions are not shared by multiple subtitle services; consequently, the page id of each page composition segment shall be equal to the composition_page_id value. In summary, the data hierarchy is: Transport Stream (TS); transport packets with the same PID; PES packets, with PTSs providing timing information; subtitle service; segments signalled by the composition page id and optionally the ancillary page id; where appropriate, a display definition segment;

14 14 subtitle data, containing information on page composition, region composition, CLUTs, objects and end of display set. 4.3 Temporal hierarchy and terminology At the segment level in the data hierarchy there is also a temporal hierarchy. The highest level is the epoch; in an epoch the page composition and the region composition may change - for example objects and regions may be added or removed. The concept of an epoch is analogous to that of an MPEG video sequence. No decoder state is preserved from one epoch to the next. An epoch is a sequence of one or more page instances. Each page instance is a completed screen of graphics. Consecutive page instances may differ little (e.g. by a single word when stenographic subtitling is being used) or may be completely different. The set of segments needed to decode a new page instance is called a display set. Within a display set the sequence of segments (when present) is: display definition segment; page composition; region composition; CLUT definition; object data; end of display set segment. All segments signalled by the composition page id value shall be delivered before any segment signalled by the ancillary page id value. The ancillary page id value shall not signal page composition segments and region composition segments. 5 Subtitle decoder model The subtitle decoder model is an abstraction of the processing required for the decoding of a subtitle service within a subtitle stream. The main purpose of this model is to define requirements for compliant subtitling streams. Figure 3 shows the prototypical model of a subtitling decoder. R1 kbit/s Subtitle Decoder R3 kbit/s MPEG 2 TS packets PID filter Transport buffer B1 bytes Pre-processor and filters Coded data buffer B2 kbytes Subtitle processing Composition buffer Pixel buffer B3 kbytes 4 kbytes Figure 3: Subtitle decoder model The input to the subtitle decoder model is an MPEG-2 Transport Stream (TS). After a selection process based on PID value, complete MPEG-2 Transport Stream packets containing the subtitle stream enter a transport buffer with a size of B1 bytes. When there is data in the transport buffer, data is removed from this buffer at a rate of R1 kbit/s. When no data is present, this data rate equals zero. For legacy decoders designed in accordance with EN (V1.2.1) [5] the transport buffer has a size of 512 bytes and the outflow rate is 192 kbits/s. For decoders capable of dealing with streams which include a display definition segment the transport buffer has a size of bytes and the outflow rate is 400 kbits/s.

15 15 The transport packets from the transport buffer are processed by stripping off the headers of the transport packets and of the PES packets. The Presentation Time Stamp (PTS) values are passed on to the next stages of the subtitling processing. In the pre-processor, the segments required for the selected subtitle service are filtered from the subtitle stream. Hence, the output of the pre-processor is a stream of subtitling segments which are filtered based on the page_id values signalled in the subtitling descriptor. The selected segments enter into a coded data buffer which has a size of B2 kbytes. For legacy decoders designed in accordance with EN (V1.2.1) [5] the coded data buffer has a size of 24 kbytes. For decoders capable of dealing with streams which include a display definition segment the coded data buffer has a size of 100 kbytes. Only complete segments are removed from this buffer by the subtitle decoder. The removal and decoding of the segments is instantaneous (i.e. it takes zero time). If a segment produces pixel data, the subtitle decoder stops removing segments from the coded data buffer until all pixels have been transferred to the pixel buffer. The pixel data of objects that are used more than once, is transferred separately for each use. The data rate for the transport of pixel data into the pixel buffer is R3 kbit/s and the pixel buffer size B3 kbytes. For legacy decoders designed in accordance with EN (V1.2.1) [5] the data rate of pixel data into the pixel buffer is 512 kbits/s and the pixel buffer size 80 kbytes. For decoders capable of dealing with streams which include a display definition segment the data rate of pixel data into the pixel buffer is 2 Mb/s and the pixel buffer is 320 kbytes. The data needed for the composition of the subtitles, such as the page composition, the region composition and the CLUTs are stored in the composition buffer, which has a size of 4 kbytes. 5.1 Decoder temporal model The requirements for memory usage in the subtitle decoder model depends on the size and colour depth of the applied regions in the page. A complete description of the memory usage of the decoder shall be delivered at the start of each epoch. Hence, epoch boundaries provide guaranteed service acquisition points. Epoch boundaries are signalled by page composition segments with a page state of type "mode change". The pixel buffer and the composition buffer hold the state of the subtitling decoder. The epoch for which this state is defined is between Page Composition Segments (PCSs) with page state of "mode change". When a PCS with page state of type "mode change" is received by a decoder all memory allocations implied by previous segments are discarded i.e. the decoder state is reset. All the regions to be used in an epoch shall be introduced by the Region Composition Segments (RCSs) in the display set that accompanies the PCS with page state of "mode change" (i.e. the first display set of the epoch). This requirement allows a decoder to plan all of its pixel buffer allocations before any object data is written to the buffers. Similarly, all of the CLUT entries to be used during the epoch shall be introduced in this first display set. Subsequent segments can modify the values held in the pixel buffer and composition buffer but may not alter the quantity of memory required Service acquisition The other allowed values of page state are "acquisition point" and "normal case". Each such "acquisition point" and "normal case" results in a new page instance. The "acquisition point" state (like the "mode change" state) indicates that a complete description of the memory use of the decoder is being broadcast. However, the memory usage shall not change. Decoders that have already acquired the service are only required to look for development of the page (e.g. new objects to be displayed). Re-decoding of previously received segments is optional. Decoders trying to acquire the service can treat a page state of "acquisition point" as if it were "mode change". Use of the page state of "mode change" may require the decoder to remove the graphic display for a short period while the decoder reallocates its memory use. The "acquisition point" state should not cause any disruption of the display. Hence it is expected that the "mode change" state will be used infrequently (e.g. at the start of a program, or when there are significant changes in the graphic display). The "acquisition point" state will be used every few seconds to enable rapid service acquisition by decoders. A page state of "normal case" indicates that the set of RCS may not be complete (the set is only required to include the regions whose region data structures - bitmap or CLUT family - are to be modified in this display set). There is no requirement on decoders to attempt service acquisition at a "normal case" display set. A display set is not required to contain a page composition segment. Within the same page composition for example a region composition may change. If no page composition segment is contained, the page state is not signalled; however, such display set will result in a new page instance equivalent to a "page update".

16 Presentation Time Stamps (PTS) Subtitling segments are encapsulated in PES packets, partly because of their capability to carry a Presentation Time Stamp (PTS) for the subtitling data. Unlike video pictures, subtitles have no innate refresh rate. Therefore all subtitle data are associated with a PTS to control when the decoded subtitle is displayed. Each PES header shall carry a PTS, associated with all the subtitle data contained within that PES packet. Consequently, for any subtitling service there can be at most one display set in each PES packet. However, the PES packet can contain concurrent display sets for a number of different subtitle services, all sharing the same presentation time. It is possible that segments sharing the same PTS have to be split over more than one PES packet (e.g. because of the 64 kbytes limit on PES packet length). In this case more than one PES packet will have the same PTS value. Subtitling segments should not be fragmented across PES boundaries. In summary, all of the segments of a single display set shall be carried in one (or more) PES packets that have the same PTS value. For each subtitling service all data of a display set shall be delivered within the constraints defined for the subtitle decoder model, so as to allow practical decoders sufficient headroom to present the decoded data by the time indicated by the PTS. NOTE: There may be times when, due for example to slightly late arrival of a complete display set or to slow rendering in the decoder, the correct time to present a subtitle (i.e. when PTS= local system clock derived from the PCR) has passed. (Late arrival can result from injudicious throttling of the bit-rate assigned to a subtitling stream at some point in the distribution network.) Under such conditions decoder designers should recognize that it is almost always better to display a late subtitle than to discard it Display definition If present in the stream, the Display Definition Segment (DDS) defines the display width and height of the TV image into which the associated DVB subtitles are to be rendered (e.g. in the case of HDTV images into by 1 080, into by 1 080, into by 720, etc.). The DDS applies to the subtitle display set being signalled and thus, if present, is transmitted once per display set. Absence of a DDS in a stream implies that the stream is coded in accordance with EN (V1.2.1) [5] and that a display width of 720 pixels and a display height of 576 lines may be assumed. The DDS also optionally signals a window within the image display into which DVB subtitles are to be rendered. This facilitates the application to HDTV services of DVB subtitles rendered for standard definition (e.g. for simulcasting SD and HDTV). Thus DVB subtitles rendered for a 720 by 576 image can be positioned within the HDTV image in a flexible manner to suit the service provider (e.g. centred horizontally and at the bottom of the HDTV frame). Annex B provides examples of how the DDS might be used in practice Page composition The Page Composition Segment (PCS) carries a list of zero or more regions. This list defines the set of regions that will be displayed at the time defined by the associated PTS. In the subtitle decoder model, the display instantly switches from any previously existing set of visible regions to the newly defined set. The PCS may be followed by zero or more Region Composition Segments (RCS). The region list in the PCS may be quite different from the set of RCSs that follow, in particular when some of the regions are initially not displayed. The PCS provides the list of regions with their spatial positions on the screen or for streams which include a display definition segment their spatial positions relative to the display window. The vertical position of the regions shall be defined such that regions do not share any horizontal scan lines on the screen. A region therefore monopolizes any part of the scan lines that it occupies; no two regions can be presented horizontally next to each other Region composition A complete set of Region Composition Segments (RCS) shall be present in the display set that follows a PCS with page state of "mode change" or "acquisition point" as this is the process that introduces regions and allocates memory for them. Display sets that represent a page update are only required to contain the data to be modified.

17 17 Once introduced the memory "foot print" of a region shall remain fixed for the remainder of the epoch. Therefore the following attributes of the region shall not change within an epoch: width; height; depth; region_level_of_compatibility; CLUT_id. Other attributes of the region specified in the RCS are the region_fill_flag and the region_n-bit_pixel_code, specifying the background colour of the region. When the region_fill_flag is set the first graphics operation performed on a region should be to colour all pixels in the region with the colour indicated by the region_n-bit_pixel_code. The value of the region_n-bit_pixel_code shall not change in RCS where the region_fill_flag is not set. This allows decoders that have already acquired the subtitling service to ignore the region_n-bit_pixel_code when the region_fill_flag is not set. A decoder in the process of acquiring the service can rely on the region_n-bit_pixel_code being the current region fill colour regardless of the state of region_fill_flag. There is no requirement for a region to be initialized by filling it with the background colour when the region is introduced at the start of the epoch. This allows the rendering load to be deferred until the region is included in the region list of the PCS, indicating that presentation of the region is required. In the limiting case, the region need never be filled with the background colour. For example, if the region is completely covered with objects Points to note At the start of an epoch the display set shall include a complete set of RCSs for all the regions that will be used during that epoch. The PCS shall only list the subset of those regions that presented at the start of the epoch. In the limiting case any PCS may list zero visible regions. An RCS shall be present in a display set if the region is to be modified. However, the RCS is not required to be in the PCS region list. This allows regions to be modified while they are not visible. An RCS may be present in a display set even if they are not being modified. For example, a broadcaster may choose to broadcast a complete list of RCSs in every display set. A decoder shall inspect every RCS in the display set to determine if the region is to be modified, for example, which pixel buffer modifications are required or where there is a modification to the associated CLUT family. It is sufficient for the decoder to inspect the RCS version number to determine if a region requires modification. There are three possible causes of modification, any or all of which may cause the modification: - region fill flag set; - CLUT contents modification; - a non-zero length object list. 5.2 Buffer memory model A page composition segment with the page state of type "mode change" destroys all previous pixel buffer and composition buffer allocations by erasing the contents of the buffers. Various processes, as detailed in the following clauses, allocate memory from the pixel and composition buffers. These allocations persist until the next page composition segment with page state of type "mode change". There is no mechanism to partially re-allocate memory within an epoch. During an epoch, the memory allocation in the pixel buffer remains the same.

18 Pixel buffer memory The pixel buffer in the subtitle decoder has a size of 80 kbytes (320 kbytes for decoders capable of dealing with streams which include a display definition segment). The pixel buffer shall never overflow. Up to 75 % is assigned for active display. The remaining capacity is assigned for future display. The subtitle decoder model assumes that all regions used during an epoch are stored in the pixel buffer and defines the following memory allocation requirement for a region in the pixel buffer: region_bits = region_width region_height region_depth where region_depth is the region's pixel depth in bits specified in the RCS. A practical implementation of a subtitle decoder may require more memory to store each region. Any such implementation dependent overhead is not taken into account by the subtitle decoder model. During an epoch, the occupancy of the pixel buffer is the sum of the region_bits of all regions used in that epoch Region memory The pixel buffer memory for a region is allocated at the start of an epoch. This memory allocation is retained until a page composition segment with page state of "mode change" destroys all memory allocations Composition buffer memory The composition buffer contains all information on page composition, region composition and CLUT definition. The number of bytes defined by the subtitle decoder model for composition buffer memory allocation is given below: Page composition except region list 4 bytes - per included region 6 bytes Region composition except object list 12 bytes - per included object 8 bytes CLUT definition excluding entries 4 bytes - per non full range entry 4 bytes - per full range entry 6 bytes 5.3 Cumulative display construction During an epoch the region modifications defined in display sets accumulate in the pixel buffer, but without any impact on the memory allocation for each region. 5.4 Decoder rendering bandwidth model The rendering bandwidth into the pixel buffer is specified as 512 kbit/s (2 Mb/s for decoders capable of dealing with streams which include a display definition segment). The subtitle decoder model assumes 100 % efficient memory operations. So, when 10 pixel 10 pixel object is rendered in a region with a 4-bit pixel depth, 400-bit operations are consumed.

19 19 The rendering bandwidth budget comprises all modifications to the pixel buffer. Certain decoder architectures may require a different number of memory operations. For example, certain architectures may require read, modify, write operation on several bytes to modify a single pixel. These implementation dependent issues are beyond the scope of the subtitle decoder model and are to be compensated for by the decoder designer Page erasure A page erasure occurs at a page time-out. Page erasure does not imply any modifications to the pixel buffer. So, page erasure does not impact rendering in the subtitle decoder model Region move or change in visibility Regions can be repositioned by altering the specification of their position in the region list in the PCS. The computational load for doing this may vary greatly depending on the implementation of the graphics system. However, the subtitle decoder model is region based so the model assumes no rendering burden associated with a region move. Similarly, the visibility of a region can be changed by including it in or excluding it from the PCS region list. As above, the subtitle decoder model assumes that no rendering is associated with modifying the PCS region list Region fill Setting the region fill flag instructs that the region is to be completely re-drawn with the defined fill colour. For example, filling a 128 pixel 100 pixel 4-bit deep region will consume bit operations, which will take 0,1 s with a rendering bandwidth of 512 kbit/s. Where the region fill flag is set, the region fill in the subtitle decoder model happens before any objects are rendered into the region. Regions are only filled when the region fill flag is set. There is no fill operation when a region is introduced at the start of an epoch. This allows the encoder to defer the fill operation, and hence the rendering burden until later. A decoder can optionally look at the intersection between the objects in the region's object list and the area to be filled and then only fill the area not covered by objects. Decoders should take into account that objects can have a ragged right hand edge and can contain transparent holes. Any such optimization is beyond the scope of the subtitle decoder model CLUT modification Once introduced a region is always bound to a particular CLUT. However, new definitions of the CLUT may be broadcast, i.e. the mapping between pixel code and displayed colour can be redefined. No rendering burden is assumed when CLUT definitions change Graphic object decoding Graphical objects shall be rendered into the pixel buffer as they are decoded. One object may be referenced several times, for example, a character used several times in a piece of text. Within a region the rendering burden for each object is derived from: the number of pixels enclosed within the smallest rectangle that can enclose the object; the pixel depth of the region where the object is positioned; the number of times the object is positioned in the region. The "smallest enclosing rectangle" rule is used to simplify calculations and also to give some consideration for the read-modify-write nature of pixel rendering processes. The object coding allows a ragged right edge to objects. No coded information is provided for the pixel positions between the "end of object line code" and the "smallest enclosing rectangle" and therefore these pixels should be left unmodified by the rendering process.

20 20 The same rendering burden is assumed, regardless of whether an object has the non_modifying_colour_ flag set to implement holes in the object. Again this gives some consideration for the read-modify-write nature of pixel rendering processes Character object decoding The subtitling system allows character references to be delivered as an alternative to graphical objects. The information inside the subtitling stream is not sufficient to make such a character coded system work reliably. A local agreement between broadcasters and equipment manufacturers may be an appropriate way to ensure reliable operation of character coded subtitles. A local agreement would probably define the characteristics of the font (character size and other metrics). It should also define a model for rendering of the characters. 6 PES packet format For carriage of DVB subtitles the PES packet syntax and semantics as defined in ISO/IEC [1] are applied within the constraints in table 1. Table 1 stream_id PES_packet_length data_alignment_indicator Presentation_Time_Stamp of subtitle PES_packet_data_byte Set to ' ' indicating "private_stream_1". Set to a value that specifies the length of the PES packet, as defined in ISO/IEC [1]. Set to '1' indicating that the subtitle segments are aligned with the PES packets. The PTS, indicating the time at which the presentation begins of the display set carried by the PES packet(s) with this PTS. The PTSs of subsequent displays shall differ by more than one video frame. The PES_data_field specified in clause 7 of the present document. 7 The PES packet data for subtitling 7.1 Syntax and semantics of the PES data field for subtitling When carrying a DVB subtitle stream, the PES_packet_data_bytes shall be encoded as the PES_data_field defined in the table below. Syntax Size Type PES_data_field() { data_identifier 8 bslbf subtitle_stream_id 8 bslbf while nextbits() == ' ' { Subtitling_segment() end_of_pes_data_field_marker 8 bslbf Semantics: data_identifier: For DVB subtitle streams the data_identifier field shall be coded with the value 0x20. subtitle_stream_id: This identifies the subtitle stream in this PES packet. A DVB subtitling stream shall be identified by the value 0x00. end_of_pes_data_field_marker: An 8-bit field with fixed contents ' '.